I must use the $\epsilon-N$ definition to prove $$\lim_{n\to\infty}\dfrac{a_1+a_2 \hspace{3px}+\hspace{3px}...\hspace{3px}+\hspace{3px}a_n}{n} = L$$ when $\displaystyle \lim_{n\to \infty} a_n = L$. Does anybody have any pointers as to how to start this problem? I'm stumped.
-
1Should the final term in the numerator have been $a_n$? – J.G. Jun 03 '19 at 11:37
-
Yes it should be $a_n$. Apologies for that. – user208480 Jun 03 '19 at 11:53
3 Answers
Let $\epsilon >0$ and choose $n_0$ such that $|a_n-L| <\epsilon$ for $n >n_0$. Let $n_1$ be an integer exceeding $\frac 2 {\epsilon\sum_{k\leq n_0} |a_k-L|}$. Then, for any $n \geq \max \{n_0,n_1\}$ we have $|\frac {a_1+a_2+..+a_n} n -L|=|\frac {(a_1-L)+(a_2-L)+..+(a_n-L)} n| < \epsilon /2+\frac 1 n (n-n_0+1)\epsilon /2<\epsilon$.

- 311,013
You know that $$L=\lim_{n\to\infty}\frac{a_{n+1}}{1}=\lim_{n\to\infty}\frac{\sum_{k=1}^{n+1}a_k-\sum_{k=1}^na_k}{n+1-n}.$$By the Stolz–Cesàro theorem (for an $\epsilon$-$N$ proof of it you can adapt to the special case of your problem viz. $b_n=n$, see here), $L=\lim_{n\to\infty}\frac{\sum_{k=1}^na_k}{n}$ as required.

- 115,835
WLOG, assume $L=0$ (otherwise, subtract $L$ from all terms).
From the hypothesis,
$$\forall\epsilon>0:\exists N:\forall n>N:-\epsilon<a_n<\epsilon.$$
Then, adding all inequalities between $N$ and $n$,
$$\forall\epsilon>0:\exists N:\forall n>N:-(n-N)\epsilon<\sum_{k=N+1}^na_k<(n-N)\epsilon,$$
which we can write
$$\forall\epsilon>0:\exists N:\forall n>N:\frac1n\sum_{k=1}^Na_k-\frac{n-N}n\epsilon<\frac1n\sum_{k=1}^na_k<\frac1n\sum_{k=1}^Na_k+\frac{n-N}n\epsilon.$$
As $n$ is unbounded, we can find an $M\ge N$ such that $\forall n>M$,
$$\frac1n\left(\sum_{k=1}^na_k+N\epsilon\right)<\frac\epsilon2$$ and we now have
$$\forall\epsilon>0:\exists M:\forall n>M:-\frac\epsilon2<\frac1n\sum_{k=1}^na_k<\frac32\epsilon.$$
-
I am having trouble understanding this answer. How do you find the second line? – user208480 Jun 03 '19 at 12:59
-
@user208480 As written, adding all inequalities between $N$ and $n$.
In a nutshell, as all terms but a finite number are close to $L$, the mean is also close to $L$ when the initial terms have been amortized.
– Jun 03 '19 at 13:03